For my research, I want to determine the sample size for unmarried male (25+ years) for a district in India. Since my study is based on rural region for the district, so what is the method to calculate this sample size from Census data (census of India, 2011).
Sorry, but the question is not really clear.
Sample size estimations are performed to ensure the power of a hypothesis test for 2 or more samples in order to detect a difference for variables of interest.
I guess, this is not what you want?!
Regards,
GL
QUESTION IS REALLY NOT CLEAR
In addition to Gunter's answer, just to elaborate, sample size estimations are also performed to limit the error involved in estimations. So, you should be clear what the estimation or comparison problem you have? What is the variable that you want to estimate or compare? Is it qualitative or quantitative? If qualitative, what is the expected proportion? If quantitative, what is the expected mean and standard deviation (SD)? If estimation problem, how much error in you estimate do you wish to allow? If comparison problem, how much power do you want to ensure. Level of confidence, Level of significance, etc?
Ambarish -
It sounds like you might be doing a quantitative study using randomized sampling. If that is correct, this still depends upon the type of data collected, sample design (stratification often helps, or division into subpopulations if you want to estimate for each), and population standard deviation by data group identified.
The finite population sampling I have in mind may not exactly be appropriate for whatever type of data you seek, however. I think you may be referring to a subject matter with other considerations, but perhaps if so, then some of this may still be analogous enough to your situation to give you a few ideas.
There is a great deal of information on finite population sampling and sample size requirements in books such as
Cochran, W.G(1977), Sampling Techniques, 3rd ed., John Wiley & Sons, if that is appropriate for you.
Note the possible desirable use of finite population correction (fpc) factors.
For more complex designs, sample size "formulas" can be messy. If you are looking at rural areas, especially if you need to travel, "cluster sampling" may be helpful. If you use stratification and cluster sampling at different stages, it does get complicated. Behind it all, however is the fact that population standard deviations are fixed but unknown, and are therefore 'guessed.' Larger sample sizes with these fixed standard deviations make standard errors go down. The idea is to sample a large enough sample to drive standard errors down to acceptable levels. Each quantitative question - say for continuous data or for yes/no data - on your survey will have different standard deviation, so a given sample size may be enough for some variables, but not others. If you need a sample size which is too large for your resources, you could bias results and increase variance with nonsampling error such as measurement error. If you do not identify your population of interest properly when collecting your randomized sample, frame errors may occur which also increase nonsampling error. But if you can do the survey without large problems of the sort, you may be able to get a good idea of standard deviations and sample size requirements by starting with a small pilot study.
Besides Cochran(1977), and among many other textbooks, depending upon your interests, here are a couple more textbooks that you may find of interest:
Blair, E. and Blair, J(2015), Applied Survey Sampling, Sage Publications.
Lohr, S.L(2010), Sampling: Design and Analysis, 2nd ed., Brooks/Cole.
[If you have additional information (auxiliary data that can be used as regressor data, i.e. for an independent variable or variables), then you could make great improvements looking into model-assisted design-based sampling methods, or even strictly model-based ones. "Small area estimation" might even be attractive, but I expect that this is out of scope here.]
So, remember all is standard deviation and design driven; check out a book such as Cochran; and consider pilot studies.
If I misunderstood, and you are looking at something else, or are not looking at quantitative data of the sort indicated above at all, you may still want to consider a pilot study, if feasible, to work out various details, including sample size needs.
Cheers - Jim
- Badges
- Science topic
- Similar topics
- Medicine
- Epidemiology
- Age Groups